Set Reorder Points for Intermittent Undercarriage Demand
Share
A shelf count cannot tell you when to reorder an undercarriage part. Some units may already be allocated, a late purchase order may be shown as incoming, and months of zero issues may be followed by several repairs at once. A useful trigger has to connect the real inventory position with demand during the entire replenishment lead time.
This guide builds that decision record for one SKU at one stocking location. It does not choose an order quantity, promise a service level or prescribe one forecasting model. First use the existing guide to choosing which undercarriage parts to stock to define the assortment. Then apply the process below to a part that the fleet has decided to hold.
Use inventory position rather than shelf quantity alone
Define the inventory state before comparing it with any reorder point. For a basic review, inventory position is usable on-hand stock plus supply that can legitimately be counted, minus backorders and other committed demand. The exact fields depend on the organization’s ERP rules, but every term should have a part number, location, unit of measure and effective date.
Start with usable stock. Separate unrestricted units from quarantine, inspection hold, damaged material, unidentified returns and stock reserved for another legal entity or location. A physical unit is not necessarily available to cover the next failure. Reconcile unusual balances with the warehouse record instead of silently adding them to the calculation.
Committed stock
Committed stock includes units already allocated to released work orders, approved repairs, customer orders or another defined demand source. Record both the quantity and the commitment identifier. This prevents the same roller, idler or track component from appearing available to two planners.
Reservations also need aging rules. A cancelled repair can release stock, while an old allocation with no owner can understate availability indefinitely. Do not delete it merely to improve the number. Route it to the person authorized to confirm or release the commitment, and retain the decision date.
Open orders and backorders
Review each open purchase order by confirmed quantity, supplier acknowledgment, expected dispatch, transport state and expected usable-receipt date. A requisition, draft PO or unconfirmed promise is not equivalent to a shipment that is likely to arrive before the exposure window closes. Define which statuses can enter inventory position and flag the rest as uncertain supply.
Subtract backorders and other unmet demand under the organization’s rule. Check for duplicate demand between a backorder, a work order and a manual shortage list. The result should be reproducible from current transactions, not a number maintained in an isolated spreadsheet.
Check the lead time the stock must cover
A reorder trigger covers demand from the decision point until replacement stock becomes usable. Measure that interval end to end: internal review and approval, supplier acknowledgment, production queue, manufacturing, inspection or document release, dispatch, transport, border handling where applicable, receiving inspection and inventory release. A quoted production duration may cover only one segment.
Build an event history from actual orders. For every suitable receipt, retain the request, PO, acknowledgment, dispatch, arrival and usable-release dates. Mark exceptional routes such as air freight or a one-time stock transfer. Mixing expedited and normal orders without labels can make ordinary replenishment appear faster than it is.
Production and transit uncertainty
Separate the sources of variation. Supplier queue time can move with capacity or raw material; transit can move with routing and border events; receiving can move with missing documents or inspection holds. Averages can hide a long tail, so review the distribution and the late cases as well as a central value.
Use the lead-time basis approved for the policy and record its scope. If evidence is sparse, run labelled scenarios rather than inventing precision. For example, compare the trigger under the normal observed route and a plausible longer route, then let the inventory owner choose the service-and-cost tradeoff. This is a planning input, not a KTSU lead-time promise.
Inspect the pattern behind average demand
Keep a complete, time-ordered demand history at the SKU and location level, including periods with zero demand. Record nonzero quantities and their operational cause where known. The review should also identify lost or hidden demand: sales or issues during a stockout may be censored, emergency substitutions may appear under another part number, and cannibalized components may never reach the inventory ledger.
Spare-parts demand is a recognized intermittent-demand problem. A critical review in Omega finds that method choice depends on the demand setting rather than one technique winning everywhere. A separate review of installed-base information explains why machine population and replacement mechanisms can add information that a bare time series omits.
Long zero-demand runs
A zero is data, but it needs interpretation. It may represent a period with no failures, an idle machine population, a maintenance deferral, a stockout that prevented an issue, or an aggregation choice that moved the demand elsewhere. Preserve the zeros and their time buckets; removing them inflates the apparent demand rate.
Elapsed time since the last demand can also contain information in some settings. Research in the European Journal of Operational Research evaluates a method that conditions on elapsed time for intermittent spare parts. That finding supports examining the timing record. It does not make elapsed time alone a reorder rule for every part.
Clustered repairs
Suppose two SKUs each issued six units over twelve months. SKU A issued one unit in six separated months. SKU B issued three units during one repair campaign and three during another. Their annual and monthly average demand is identical, but a one-unit buffer experiences the events differently. The second pattern may reflect campaign maintenance, common operating conditions or several machines of the same age entering repair together.
Tag known campaigns, fleet transfers and common-cause events. Do not erase them as outliers simply because they are inconvenient. Decide whether the event can recur within the next lead time and test the trigger against both ordinary and clustered sequences.
Compare trigger policies with explicit assumptions
A transparent starting point is estimated demand during the defined replenishment lead time plus a separately approved buffer. Compare current inventory position with that provisional trigger. Keep the units and time basis consistent. If demand is stored weekly and lead time is recorded in calendar days, document the conversion and avoid false precision.
Simple planning baseline
The baseline is useful because a reviewer can see each input. It is weak when the average comes from few nonzero events, when lead time varies widely, or when sales history is censored by stockouts. Do not call the buffer “safety stock” unless its method, service objective and owner are defined. An arbitrary extra unit may be a scenario, but it is not a statistically validated protection level.
Use a review table that keeps missing and conflicting evidence visible:
| Review state | Inventory evidence | Demand and lead-time evidence | Decision |
|---|---|---|---|
| Normal | Usable stock, commitments, backorders and credible open supply reconcile | Complete zero/nonzero history and end-to-end receipt dates support the stated basis | Run the approved trigger and record the result |
| Missing | Allocation status, quarantine quantity or PO confirmation is absent | Stockout-censored demand or usable-receipt date is unknown | Keep the trigger provisional and obtain the missing record |
| Conflict | Warehouse, ERP and repair commitments disagree | Supplier promise, dispatch record and receipt history imply different lead times | Pause automation, resolve the conflict and preserve the owner’s decision |
When specialist forecasting is needed
Escalate when the consequence of shortage is high, data are sparse or censored, events are strongly clustered, lead-time demand has a heavy tail, parts are superseding, or several locations share emergency stock. Intermittent series create challenges for forecasting, model comparison and error measurement, as summarized in an open peer-reviewed review and modelling paper.
A qualified analyst can compare occurrence-and-size methods, empirical lead-time demand, simulation or other suitable approaches using walk-forward tests and inventory outcomes. Compare policies on shortage exposure, emergency actions, holding cost and aged stock, not only an average forecast-error score. Retain the simple baseline so stakeholders can see whether added complexity improves the operational decision.
Review the trigger after actual events
For every replenishment cycle, capture the inventory position and trigger at order time, order quantity, supplier acknowledgment, usable receipt, and demand that occurred before receipt. Compare the event with the assumptions. A policy review should have a named owner, effective date and next review date; otherwise parameter changes become untraceable.
Stockouts
Diagnose a stockout before raising the trigger. Was inventory position wrong because an allocation or hold was missing? Did demand exceed the tested event pattern? Did supply arrive late, or did receiving delay release? Estimate unmet demand carefully because recorded issues stop when no stock is available. Correct the cause that evidence supports.
Emergency orders
Record why an expedite, transfer or substitute was used and what it cost. An emergency order can reveal a policy weakness, but it may instead be the response to a one-time failure cluster or supplier delay. Keep exceptional lead time separate from the normal-route history while retaining it for disruption scenarios.
Changed machine population
Update the input when machines enter or leave service, utilization changes, a maintenance campaign starts, or a part number is superseded. Installed-base evidence can explain demand changes that historical averages cannot. Confirm compatibility and supersession separately; a new machine count does not prove a proportional demand rate.
The final record should show the current inventory-position rule, lead-time basis, demand data window, scenario or model, approved buffer, trigger, owner and unresolved evidence. Automate replenishment only after those fields reconcile and actual cycles have been reviewed. The goal is a trigger that can be challenged and improved, not a formula that hides uncertainty.